kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-02-18 08:45:05 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

[None][chore] Add warning about 2-model MTP deprecation (#11043)

Browse Source 
Signed-off-by: Mike Iovine <miovine@nvidia.com>
Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
This commit is contained in:
Mike Iovine 2026-02-02 10:35:51 -05:00 committed by Yanchao Lu
parent 5e47e6970b
commit 435ea36977
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tensorrt_llm/llmapi/llm_args.py
View File

@ -1183,6 +1183,12 @@ class MTPDecodingConfig(DecodingBaseConfig):
self.max_total_draft_tokens = kwargs[
'num_nextn_predict_layers'] # Current MTP only support linear tree
if not self.mtp_eagle_one_model:
logger.warning(
"2-model style MTP is deprecated. The mtp_eagle_one_model flag will do nothing "
"in release 1.3. After that, the flag will be removed entirely."
)
@classmethod
def from_dict(cls, data: dict):
out = cls(**data)